The Invisible Trillion

How Much Do We Actually Need?

The Question The Industry Doesn’t Want Asked

The chapters on the people next door described, at length, what is being done to people in order to build the AI data centers of the 2020s. The noise, the air, the bills, the cascade. The legal weapons and the tactical tricks. The neighborhoods entering trajectories that will, in many cases, end in collapse twenty years from now.

There is a question I have not yet asked, and it is the question the entire rest of this book is dancing around. The question is this:

How much of this do we actually need?

If the buildout described in the neighborhood chapters is paying for compute capacity that will be used to do things people genuinely want done — to cure diseases, to make scientific discoveries, to translate languages, to help children learn, to do useful work that nobody could do before — then the moral situation, while still containing real injustices that should be addressed, is at least one where there is a real benefit being created somewhere. The harm to the neighborhoods is being paid for something. We can argue about whether the something is worth the cost, but at least the something exists.

If, on the other hand, the buildout is paying for compute capacity that will mostly be used to produce search-engine spam, automated marketing copy, content nobody reads, ads nobody clicks on, and AI-generated images that fill social media feeds and are scrolled past without engagement — then the moral situation is much worse. The neighborhoods are being destroyed to produce junk. The wealth transfer described in those chapters is happening in service of an industry whose actual output, at the scale being built, has very little value to anyone except the companies producing it.

These are different worlds, and which one we are actually in matters more than almost any other question this book can ask. The two cases look identical from outside — the same data centers, the same power draws, the same neighborhoods hosting them — right up until the demand either materializes or doesn’t. And those costs are imposed before anyone knows which way it goes.

Here is the spoiler. The evidence available in mid-2026 strongly suggests we are not in the first world but something closer to the second, with some genuine demand mixed in, and the buildout is running well ahead of any rigorous demand signal. The industry knows this. Its investors are starting to. The residents of those neighborhoods mostly do not, because nobody has told them.

All 19 chapters of The Invisible Trillion · Library · All courses